Journal of Clinical Medicine Article Genome-Wide Scan for Copy Number Alteration Association with Relapse-Free Survival in Colorectal Cancer with Liver Metastasis Patients Po-Sheng Yang 1,2 , Hsi-Hsien Hsu 3, Tzu-Chi Hsu 3, Ming-Jen Chen 3, Cin-Di Wang 4, Sung-Liang Yu 5, Yi-Chiung Hsu 6,* and Ker-Chau Li 4,7 1 Department of Medicine, Mackay Medical College, New Taipei 252, Taiwan; [email protected] 2 Department of General Surgery, Mackay Memorial Hospital, Taipei 104, Taiwan 3 Department of Colorectal Surgery, Mackay Memorial Hospital, Taipei 104, Taiwan; [email protected] (H.-H.H.); [email protected] (T.-C.H.); [email protected] (M.-J.C.) 4 Institute of Statistical Science, Academia Sinica, Taipei 115, Taiwan; [email protected] (C.-D.W.); [email protected] (K.-C.L.) 5 Department of Clinical Laboratory Sciences and Medical Biotechnology, College of Medicine, National Taiwan University, Taipei 100, Taiwan; [email protected] 6 Department of Biomedical Sciences and Engineering, National Central University, Taoyuan 320, Taiwan 7 Department of Statistics, University of California Los Angeles, Los Angeles, CA 90095, USA * Correspondence: [email protected]; Tel.: +886-3-4227151 (ext. 27752) Received: 23 October 2018; Accepted: 13 November 2018; Published: 18 November 2018 Abstract: Predicting a patient’s risk of recurrence after the resection of liver metastases from colorectal cancer is critical for evaluating and selecting therapeutic approaches. Clinical and pathologic parameters have shown limited accuracy thus far. Therefore, we combined the clinical status with a genomic approach to stratify relapse-free survival in colorectal cancer liver metastases patients. To identify new molecular and genetic signatures specific to colorectal cancer with liver metastasis (CRCLM) patients, we conducted DNA copy number profiling on a cohort of 21 Taiwanese CRCLM patients using a comparative genomic hybridization (CGH) array. We identified a three-gene signature based on differential copy number alteration between patients with different statuses of (1) recurrence and (2) synchronous metastasis. In relapse hotspot regions, only three genes (S100PBP, CSMD2, and TGFBI) were significantly associated with the synchronous liver metastasis factor. A final set of three genes—S100PBP, CSMD2, TGFBI—significantly predicted relapse-free survival in our cohort (p = 0.04) and another CRCLM cohort (p = 0.02). This three-gene signature is the first genomic signature validated for relapse-free survival in post-hepatectomy CRCLM patients. Our three-gene signature was developed using a whole-genome CGH array and has a good prognostic position for the relapse-free survival of CRCLM patients after hepatectomy. Keywords: colorectal cancer liver metastases; copy number alteration; gene signature; relapse-free survival; biomarker 1. Introduction Colorectal cancer (CRC) is the third most common human malignancy worldwide and the third leading cause of cancer death in the U.S. [1,2]. In Taiwan, the number of new cases of CRC has increased every year, becoming the most and second most prevalent form of cancer in males and females, respectively [3]. Despite the improvement of early diagnostics, synchronic metastasis was noted in about 20%–25% of CRC patients at diagnosis, and these patients had a less than 10% 5-year survival rate [4]. About 50% of CRC patients developed liver metastasis after treatment of their primary J. Clin. Med. 2018, 7, 446; doi:10.3390/jcm7110446 www.mdpi.com/journal/jcm J. Clin. Med. 2018, 7, 446 2 of 13 tumors, and approximately one-third of these metachronous patients had the disease confined to the liver [5,6]. The median survival was about 5–10 months for CRC patients with liver metastasis without treatment, and less than 0.5% of these patients survived beyond 5 years [7]. For colorectal cancer with liver metastasis (CRCLM), hepatic resection remains the only option with the potential to cure. However, only 15%–25% of CRCLM patients are cured, and 70% experience tumor recurrence [8,9]. Perioperative systemic therapy is usually suggested in patients with resected CRCLM, but a large randomized controlled trial showed that there was no improvement in 5-year overall survival (OS) compared to patients treated with hepatic resection alone (51% vs. 48%) [10]. Almost 30% of patients died with cancer within 2 years after surgery for CRCLM [10], and the selection of optimal treatments for metastatic colorectal cancer is still a complex issue. Therefore, the development of the new molecular and genetic signatures to identify patients at a high risk of relapse after hepatectomy for CRCLM is important. Many studies have shown that DNA copy number alteration (CNA) correlates with outcome in colon cancer patients [11–14], and somatic CNA is crucial for the development of CRC [15]. Tumor metastasis is a complex process, and the series of molecular events leading to metastasis is still unclear [16–20]. While several studies have focused on genetic heterogeneity in the many primary malignancies of CRCs, there is consensus on the genetic heterogeneity between a primary cancer and its distant metastasis [21–23]. Furthermore, no prognostic CNA signatures have been developed to assess outcomes after hepatectomy for CRCLM. To identify new molecular and genetic signatures specific to CRCLM patients, we conducted DNA copy number profiling on a cohort of 21 Taiwanese CRCLM patients by a comparative genomic hybridization (CGH) array. We identified a three-gene signature associated with cancer recurrence which was prognostic for relapse-free survival in our cohort. We externally validated this signature in a public cohort of 45 patients after hepatectomy for CRCLM. Herein, we identify a three-gene signature that is prognostic for relapse-free survival and present the validation results of this signature in an independent cohort. 2. Materials and Methods 2.1. Patient Samples We obtained 21 paired normal liver and CRCLM metastatic tissues of patients from the anonymized specimens deposited in the Mackey Memorial Hospital tissue bank, in accordance with the protocol approved by the Mackey Memorial Hospital’s Institutional Review Board (13MMHIS009). 2.2. Array CGH and Data Processing The whole-genome NimbleGen CGH array (NimbleGen®; NimbleGen Systems Inc, Madison, WI, USA) containing 385,806 probes with spacing of around 6000 bp was used for comparative genomic hybridization of DNA from frozen cancer tissues compared to normal DNA extracted from the PBMC of one male and one female in a community cohort. Patient DNA was required to pass a quality check by agarose electrophoresis. Digital sonifier (Branson Model#450, Branson, Danbury, CT, USA) was used for DNA fragmentation. Labeling, hybridization, and washing were performed according to the manufacturer’s protocol. Array scanning and image generation were performed by the GenePixTM Reader (Personal 4000B, Axon Instruments, Molecular Devices, Sunnyvale, CA, USA) and GenePix® Pro 6.0 software (Axon Instruments, Union City, CA, USA). Generation of log-intensity ratio data with normalization was performed by NimbleScanTM version 2.4 (Roche Nimblegen, Madison, WI, USA) and SignalMapTM version 1.9 software (Roche Nimblegen Madison, WI, USA). For signal enhancing, elementary blocks were formed from the original 385,806 probes in the array by grouping 10 consecutively located probes together [24]. Visualization of CNA profiles was performed by applying the GLAD algorithm (R package from Bioconductor, ). We computed the t-test-significant probe density and produced a density bar plot by the CIRCOS rings. The CIRCOS program was J. Clin. Med. 2018, 7, 446 3 of 13 downloaded from http://circos.ca/. The CGH array and clinical data were submitted to the Gene Expression Omnibus archive, available under accession number GEO: GSE103088. 2.3. Statistical Analysis Continuous variables are expressed as the mean ± SD or median (range) according to their homogeneity. Demographic and clinical variables were analyzed by Fisher’s exact test and multivariable logistic regression. Statistical significance was defined as a two-tailed p value < 0.05. Fisher’s exact test and logistic regression were accomplished using R version 3.3.1 (https://www.r- project.org/). 2.4. Functional Enrichment Analysis of Synchronous Metastasis-Associated Genes We use the Ingenuity Pathway Analysis (IPA) (QIAGEN company, Redwood City, CA, USA), a web-based computational platform, to conduct functional enrichment analysis of genes. We input the set of 119 synchronous metastasis-associated genes (273 probes) and used the Core analysis enrichment tool with the default settings. 2.5. Validation Cohort Genome-wide copy number analysis was performed on 45 patients with metastatic colorectal cancer using the Affymetrix SNP 6.0 Array [25]. We used the copy number variation of the first metastasized part in each patient and associated it with the overall and relapse-free survival (GSE63490). 2.6. Survival Analysis A patient’s risk score was calculated as the sum of the levels of copy number variation of each gene. Patients were classified as having a high-risk gene signature or a low-risk gene signature, with the median of the risk score as the threshold value. Survival curves for both groups were obtained by the Kaplan–Meier method and were compared using the log-rank test. Both univariate and multivariate Cox regression models were applied for prediction of patient survival. Both the log-rank test and the Cox test were two-sided, and a p-value < 0.05 was considered statistically significant. 3. Results 3.1. Patient Demographics A total of 21 CRCLM patients between December 2009 and December 2011 were included in the study, comprising 12 colon and 9 rectal cancer patients. Ages ranged from 39 to 83 years old (mean 60.9 years old) with 12 males and 9 females. Follow-up times ranged from 14.5 months to 55.7 months with a mean of 30.2 months. The primary CRC tumor size ranged from 1.2 to 8 cm.